Analyzing distributed medical databases on datamininggrid©

Abstract

Hospitals throughout Europe hold vast amounts of data in the form of patient records. Performing on-the-fly analyses of these data and their actual transformation into information and knowledge may help improve medical procedures, treatments or prevent illnesses. Grid technology has recently emerged to address the needs for efficient and effective exploitation of heterogeneous and geographically distributed resources, such as large and distributed data, open source or proprietary programs for data analysis, massive storage devices and high-performance computers. A de facto standard framework for building grid environments is the Open Grid Service Architecture (OGSA) and the corresponding Web Service Resource Framework (WSRF). The Globus Toolkit version 4, is a fully WSRF-compliant grid middleware, which addresses the needs for secure, flexible, interoperable and seamless use of grid resources. The DataMiningGrid© (www.datamininggrid.org) system was recently built on top of existing Globus technology inter alia to address the requirements of a community of medical users and enable them to perform on-the-fly analysis of geographically distributed medical databases. DataMiningGrid© is a set of grid services and user-friendly workflow editing and managing tools, which facilitate manipulation of distributed data, registering, discovery and use of grid-enabled statistical and data mining programs, their execution in the grid environment and a provenance tracking mechanism. The software is now freely available at SourceForge.net under Apache License V2. The present work illustrates the use of the DataMiningGrid© system to perform analysis of nine regional medical databases in Slovenia.